Structural Characterization of B and Non-B Subtypes of HIV-Protease: Insights Into the Natural Susceptibility to Drug Resistance Development

doi:10.1016/j.jmb.2007.03.049 J. Mol. Biol. (2007) 369, 1029–1040

Structural Characterization of B and non-B Subtypes of HIV-Protease: Insights into the Natural Susceptibility to Drug Resistance Development

Mario Sanches1, Sandra Krauchenco1, Nadia H. Martins1 Alla Gustchina2, Alexander Wlodawer2 and Igor Polikarpov1⁎

1Grupo de Cristalografia, Although a majority of HIV-1 infections in Brazil are caused by the subtype B Instituto de Física de São Carlos, virus (also prevalent in the United States and Western Europe), viral Universidade de São Paulo, subtypes F and C are also found very frequently. Genomic differences Av. Trabalhador Saocarlense, between the subtypes give rise to sequence variations in the encoded 400, CEP 13560-970, proteins, including the HIV-1 protease. The current anti-HIV drugs have São Carlos, SP, Brazil been developed primarily against subtype B and the effects arising from the combination of drug-resistance mutations with the naturally existing 2Protein Structure Section, polymorphisms in non-B HIV-1 subtypes are only beginning to be Macromolecular elucidated. To gain more insights into the structure and function of different Crystallography Laboratory, variants of HIV proteases, we have determined a 2.1 Å structure of the native National Cancer Institute subtype F HIV-1 protease (PR) in complex with the protease inhibitor TL-3. at Frederick, Frederick, We have also solved crystal structures of two multi-drug resistant mutant MD 21702, USA HIV PRs in complex with TL-3, from subtype B (Bmut) carrying the primary mutations V82A and L90M, and from subtype F (Fmut) carrying the primary mutation V82A plus the secondary mutation M36I, at 1.75 Å and 2.8 Å resolution, respectively. The proteases Bmut, Fwt and Fmut exhibit sevenfold, threefold, and 54-fold resistance to TL-3, respectively. In addition, the structure of subtype B wild type HIV-PR in complex with TL-3 has been redetermined in space group P61, consistent with the other three structures. Our results show that the primary mutation V82A causes the known effect of collapsing the S1/S1′ pockets that ultimately lead to the reduced inhibitory effect of TL-3. Our results further indicate that two naturally occurring polymorphic substitutions in subtype F and other non-B HIV proteases, M36I and L89M, may lead to early development of drug resistance in patients infected with non-B HIV subtypes. © 2007 Elsevier Ltd. All rights reserved. Keywords: HIV-1 protease structures; non-B subtype; drug resistance; *Corresponding author subtype F protease; L89M mutation

Present address: M. Sanches, Brazilian Synchrotron Introduction Light Laboratory (LNLS), Center for Structural Molecular Biology, Caixa Postal 6192, CEP 13084-971, Campinas, SP, Acquired immunodeficiency syndrome (AIDS) is Brazil. a complex of symptoms and diseases resulting from Abbreviations used: HIV-PR, human immunodeficiency infection by the human immunodeficiency virus virus protease; Bwt, subtype B wild-type HIV-PR; Fwt, (HIV). HIV is a member of the Lentivirus genus, subtype F wild-type HIV-PR; Bmut, multi-drug resistant which also includes, among others, simian immu- subtype B mutant HIV-PR; Fmut, drug-resistant subtype F nodeficiency virus (SIV) and feline immunodefi- mutant HIV-PR; PI, a commercial protease inhibitor used ciency virus (FIV).1 HIV is characterized by a wide in AIDS therapy. range of viral genetic diversity among the distinct E-mail address of the corresponding author: types, groups, and clades. The two major distinct [email protected] types of HIV, HIV-1 and HIV-2, are distinguished by

0022-2836/$ - see front matter © 2007 Elsevier Ltd. All rights reserved. 1030 Structural Characterization of B and F HIV PR their genome organization and phylogenetic relation- of HIV-1 are still at the early stages. Therefore, with ship. Further analyses of different strains of HIV-1, the aim of expanding the information about non-B from diverse geographical origins, show that isolates subtypes and their relation to drug resistance, we can be subdivided into groups, subtypes, sub- present the crystal structures and kinetic studies for subtypes, and circulating recombinant forms (CRFs), three multi-drug-resistant variants of HIV-PR isolated based on phylogenetic sequence differences. Groups from Brazilian patients: a wild-type HIV-PR of refer to distinctive HIV-1 lineages, M (for Major), O subtype F (Fwt) obtained from a naïve individual (for Outlier), and N (for New, or Non-M, Non-O); a carrying polymorphic mutations I15V, E35D, M36I, majority of strains found worldwide belong to group S37A, R41K, R57K, D60E, Q61N, I62V, L63S, I64L, M. Nine subtypes of HIV-1 group M (A–D, F–H, J and L89M and two mutant HIV-PR isolated from patients K) are currently identified, as well as 14 circulating failing intensive anti-AIDS therapy. One of the latter is recombinant forms (CRFs),2 the most common of of subtype B (Bmut), with mutations S37A, R41K, which are CRF01_AE and CRF02_AG.3 The K45R, I54V, L63P, A71V, V82A, L90M, and the other sequences of these subtypes and recombinant forms of subtype F (Fmut)withmutationsL10I,I15V, differ from one another by 25–35% in the env gene, G16E, K20R, E35D, M36I, S37N, P39S, R41K, M46I, and 10–15% in the pol gene, which includes the coding G51R, I54V, R57K, D60E, Q61D, K70R, I72V, T74A, regions for protease (PR) and reverse transcriptase V82A, L89M, with respect to the B wild-type (Bwt) (RT). Although the pol gene is the most conserved sequence. An additional Q7K mutation was region of HIV-1, there is sufficient diversity to allow included in all the constructs to increase stability 4 9 phylogenetic subtype identification. toward autolysis. The C2-symmetric inhibitor TL- Almost all studies on drug susceptibility of HIV-1 310,11 was used for co-crystallization and inhibitory have been performed in developed countries, where studies. The structures of Fwt and Fmut HIV-PR subtype B still dominates the epidemic, but on the reported here are the first for any non-B subtype worldwide scale this is not a predominant HIV HIV-PRs available to date. subtype.3 The Brazilian epidemic is characterized by the presence of multiple HIV-1 group M subtypes, primarily subtype B and subtype F, but also subtypes Results and Discussion C, D, and other circulating recombinant forms.5 HIV-1 PR continues to be one of the primary targets of AIDS drug discovery due to its central role in Protease sequences processing of viral polypeptide precursors.6 Although inhibitors of HIV-PR slow down the The sequence alignment of the four HIV proteases progress of the disease, they do not completely studied by us is presented in Figure 1. Positions of suppress viral replication, and the rapid development the primary and secondary mutations with respect of drug resistance decreases their efficacy. More than to the Bwt sequence are highlighted. The subtype F 87 mutations have already been reported in at least 49 wild-type HIV-1 PR (Fwt) naturally carries the positions within the 99 residue-long HIV-PR; many of mutations I15V, E35D, M36I, S37A, R41K, R57K, them have been identified as potential contributors to D60E, Q61N, I62V, L63S, I64L, and L89M with resistance toward one or more inhibitors.7 These respect to the subtype B sequence. The mutant of B mutations are labeled as primary and secondary, with subtype HIV-PR (Bmut) contains eight mutations the primary mutations directly reducing drug sus- (S37A, R41K, K45R, I54V, L63P, A71V, V82A, and ceptibility, whereas the secondary mutations contri- L90M) and the mutant of F subtype HIV-PR (Fmut) bute to resistance by reducing drug susceptibility or includes 20 mutations (L10I, I15V, G16E, K20R, improving the replicative fitness of isolates with a E35D, M36I, S37N, P39S, R41K, M46I, G51R, I54V, primary mutation.8 The type and location of possible R57K, D60E, Q61D, K70R, I72V, T74A, V82A, and mutations is restricted by the necessity of the virus to L89M). Finally, the point mutation Q7K was in- produce an active enzyme with sufficient affinity for troduced in all four enzymes to prevent autolysis the substrate, in order to maintain viability. Most without affecting enzymatic activity.9,12 primary mutations occur in the active site and, although they usually preserve its charge and Crystallographic structure solution polarity, they alter its geometry. Other mutations cause resistance by altering enzyme catalysis, dimer All four structures of HIV-PR presented here stability, inhibitor binding kinetics, conformational were solved with the inhibitor TL-3 (Figure 2(a)) dynamics, or by reshaping the active site through bound in the active site, in order to allow for long-range structural perturbations.8 Extensive use precise structural comparison and to assess the of anti-retroviral drugs may cause different effects differences between the B and F subtypes of HIV on distinct subtypes, due to additional selective PR. TL-3 is a symmetrical inhibitor originally pressure on HIV-PR. The polymorphism of subtypes developed against FIV PR, which proved to be a could modulate anti-retroviral drug susceptibility good universal inhibitor of most retroviral pro- and the possibility of drug resistance during therapy teases, with high potency against HIV-PR.10,11 as well the fitness of HIV-1 variants.4 The choice of the space group for the hexagonal Studies involving the effects of the available HIV-PR crystals of HIV-1 PR has been a matter of some inhibitors in persons infected with different subtypes controversy in the past, with no unambiguous Structural Characterization of B and F HIV PR 1031

Figure 1. Alignment of the sequences of the four enzymes discussed in this article. The primary mutations V82A and L90M are marked in red, whereas the secondary mutations analyzed here are marked in black. The polymorphic substitution L89M is marked in gray to depict our hypothesis that L89M is a structural equivalent of L90M. Identical residues among all four sequences are marked with an asterisk on the consensus line, while those conserved in three sequences are marked with a colon. The consensus line is left blank otherwise.

solution to date. Since diffraction data can be scaled crystallographic axis of the P6122 space group, equally well in space groups P61 and P6122, the latter with two subunits per asymmetric unit. Slight can be assumed to represent the simplest description structural differences at the surface of the two non- of the contents of a crystal.13 However, the situation crystallographic symmetry (NCS)-related subunits, can be also described as perfect merohedral twinning which include distinct conformation of a number of of the former space group, or be due to the presence side-chains and of the inhibitor, as well as different of only small differences between two quasi-sym- positions of several water molecules, can only be metric molecules in the latter one.14 In practice, since described as disorder in the higher-symmetry space the structures of the HIV-PR dimers complexed with group, but can be described more simply in the inhibitors always contain some asymmetry, it is lower-symmetry one. Indeed, the high similarity easier to describe them in the lower-symmetry space between the two crystallographically independent group, regardless of the “true” space group, and that molecules can be assessed by comparing the root- led us to the approach described below. mean-square deviation (RMSD) between their α The subtype B wild-type protease in complex with equivalent C coordinates (Table 1), which explains TL-3 was earlier solved and refined in the space why the original data could have been processed in 11 group P6122. Following the previous procedures, space group P6122. we initially also processed the Bmut and Fwt As a consequence of the assignment of the P61 datasets in space group P6122 (the systematic space group, we could now more easily model a absences clearly showed the presence of the 61 or complete molecule of the TL-3 inhibitor in the active 65 axis), obtaining Rsym of 4.6% and 7.5%, res- site of HIV-1 PR. The inhibitor occupies two distinct pectively.15 Both Bmut and Fwt structures were conformations with half occupancy each, bound to successfully solved with one subunit and half the theactivesiteinanasymmetricmanner.This inhibitor molecule in the asymmetric unit. However, behavior fully explains the three lobes of electron for both structures the values of Rfree were increas- density present in the active site, with one hydroxyl ing significantly after a few cycles of refinement. of the central diol occupying the middle lobe in each After a number of frustrating attempts to satisfacto- of the two conformations (Figure 2(b)). rily refine these models, the problems were finally A peculiarity of the hexagonal packing of the PR solved by reprocessing the data in the lower- molecules is the disorder of the Phe53 side-chain in 16 symmetry space group P61,whichhasledto both subunits. Since the side-chain aromatic ring acceptable statistics for the final refined models falls on the crystallographic 6-fold screw axis, it had (Table 1). The Bwt structure was previously refined to be refined in two mutually exclusive orientations in space group P6122 with a wide spread between in order to properly account for this disorder. 11 Rfree/Rcryst (0.281 and 0.188, respectively) and with The original Bwt-TL-3 structure contains one half only 33 water molecules in the final model.11 In of the inhibitor in the asymmetric unit, with the order to improve these statistics and to provide a whole inhibitor molecule being defined by a consistent framework for all structures compared crystallographic 2-fold axis. This implies symmetric here, we reprocessed the previously collected data in interaction of the inhibitor with the active site. space group P61 and refined the structure against Furthermore, in this model the inhibitor has a flip this diffraction data set. The final, refined model has in the peptide bond between the groups Ala in P3 Rfree/Rcryst of 0.219/0.159 and contains 98 water and Cbz in P4 (Figure 2(c)). In all four structures molecules. The Fmut structure was solved from the presented here, including the Bwt protease repro- beginning in the space group P61 (Table 1). cessed in space group P61, there is no evidence of In the lower-symmetry space group P61 a non- anomalous inhibitor conformation (Figure 2(c)). crystallographic 2-fold axis replaces the 2-fold Such a conformation was most likely an artifact 1032 Structural Characterization of B and F HIV PR

10 – Figure 2. (a) Structural representation of the C2-symmetric inhibitor TL-3 . (b) The final 2Fobs Fcalc electron density contoured at 1σ, superposed with the refined model of the TL-3 inhibitor, for the Bmut structure. The three lobes of electron density, marked with numbers, are explained by the double conformation of the inhibitor. This central part of the inhibitor, which contains the non-hydrolyzable peptide bond, occupies the active site, with the diols interacting with the catalytic aspartate residues. For clarity, one conformation is colored green and the other one yellow. (c) The final – σ 2Fobs Fcalc electron density contoured at 1 for the P3 and P4 subsites of the inhibitor TL-3 bound to the Bwt PR. A model for the TL-3 inhibitor is shown in red (both conformations). The previously published Bwt structure in P6122 was superimposed on this model and is shown in green. Apparently, the flip observed in that structure is an artifact. caused by the refinement of one half of a symmetric structures are 20.1, 20.7, 24.9 and 21.1 Å2, respec- inhibitor in a single conformation, not sufficient to tively. These low B values are typical for well- model fully its interactions in the HIV-PR active ordered structures, thus allowing for precise com- site. parison and analysis of these models. The largest difference among the models is Comparison of the structures observed for the loop between residues 35 and 41. Superposition of the inhibitors bound to each of Unless explicitly stated, all the analyses and the structures shows that they interact with the comparisons performed throughout this article enzymes in a very similar fashion, with the highest refer to subunit A. As expected, the observed divergence confined to the residue Cbz in the structural differences between the four refined subsite P4/P4′, probably due to the lack of models are very small, with a positional RMSD of interactions between this residue and the S4/S4′ 0.43 Å (Bwt/Bmut), 0.58 Å (Bwt/Fwt), 0.55 Å (Bwt/ pocket of the enzyme. Fmut), 0.44 Å (Fwt/Bmut), 0.47 Å (Fwt/Fmut), and α 0.43 Å (Bmut/Fmut) for all 198 C atoms of the Differences between HIV-PR of the B and F viral dimer. The mean temperature factors for each subtypes structure are 32.8, 30.1, 36.1 and 47.5 Å2, for Bwt, Bmut, Fwt and Fmut, respectively. The mean temp- None of the polymorphic substitutions that erature factors for the TL-3 inhibitor in the same differentiate the wild-type subtype F protease Structural Characterization of B and F HIV PR 1033

Table 1. Crystallographic data and statistics for the four HIV-PRs

Bwt Bmut Fwt Fmut Cell parameter [a=b, c] (Å) 63.13, 83.41 60.93, 82.46 61.43, 80.89 61.25, 82.23 Resolution (Å) 54.72–2.10 (2.21–2.10) 52.78–1.75 (1.84–1.75) 53.23–2.10 (2.21–2.10) 53.07–2.80 (2.95–2.80) Dataset completeness (%) 99.7 (99.7) 99.6 (99.6) 99.7 (99.7) 99.9 (99.9) I/σ(I) 13.6 (2.7) 17.8 (3.4) 13.0 (3.1) 9.8 (2.0) a Rsym (%) 7.5 (41.4) 3.7 (34.8) 4.6 (31.8) 8.8 (48.2) Redundancy 3.7 (3.7) 4.8 (4.8) 2.5 (2.4) 3.2 (3.2) Rcryst/Rfree 0.159/0.219 0.183/0.232 0.194/0.265 0.183/0.260 RMSD from ideality Bond distance (Å) 0.015 0.015 0.013 0.015 Bond angle (°) 1.57 1.78 1.59 1.69 Number of water molecules in AU 98 189 117 50 RMSD between subunitsb 0.047 0.058 0.057 0.039 c Ki (nM) 3.3±0.9 11±3 24±8 180±29 PDB ID code 2P3B 2P3A 2P3C 2P3D

All structures were solved and refined in the space group P61. The values in parentheses refer to the highest resolution shell. The inhibition constants (Ki), determined for all four enzymes against TL-3 inhibitor, are also shown. a This refers to the R value reported by the program SCALA. merge α b Root-mean-square deviation between the C coordinates of the two equivalent subunits in the asymmetric unit. c – Inhibition constants (Ki) for the inhibitor TL-3 obtained at 37 °C by measuring the rate of substrate hydrolysis using 5 50 nM protease in 100 mM sodium acetate (pH 4.7), 1 M sodium chloride, 1 mM EDTA, 1 mM DTT, 10% DMSO and 10 μM substrate plus increasing amounts of inhibitor.

from wild-type subtype B involve residues that resulting in retracing of the main chain of the create the substrate-binding pockets. In fact, both loop containing this mutation, and in its displace- structures are very similar, with an overall RMSD ment toward the loop 76-83, which forms the S1/ α of only 0.58 Å between the C coordinates of S1′ pocket. In fact, in Bwt PR structure, the loop equivalent subunits. Nevertheless, significant struc- 34-40 is involved in 27 VDW interactions and tural differences are observed in the region between three H-bonds with the loop 76-83, compared to residues 33–42, with a maximum deviation of 40 VDW and five H-bonds in the Fwt PR α 2.85 Å between the C positions of residue 35 structure, implying better stabilization of the (Figure 3(a)). The residues 33 to 42 are confined to a catalytic pocket loop 76-83 in the latter enzyme. loop that can be defined as the flap hinge, a region In total, we found 18 H-bonds between the flap that exhibits extensive rearrangement between the hinge (residues 34–40) and the rest of the protein unliganded enzymes and their complexes with residues in the Bwt structure, compared to 25 in inhibitors.17 Among the differences in the sequence Fwt. In agreement with our hypothesis, displace- found in this region, only M36I is associated with ment of the flap hinge loop is not observed in the resistance to inhibitors, and is considered a subtype Bmut structure, which does not carry the M36I B secondary mutation (Figure 1) developed along mutation (Figure 3(b)). with primary mutations in individuals under The working hypothesis of the stiffening of the intensive treatment, as well as a common substitu- flap hinges for F subtype HIV-PR carrying the tion in non-B wild-type HIV proteases. Our M36I mutation is also supported by a compar- hypothesis is that the mutation M36I impairs the ison of the temperature factors (Figure 4). In Fwt movement of the hinges of the flaps. The flaps PR the magnitude of the temperature factors is (residues 33–62) extend over the substrate-binding consistently higher than in Bwt PR, an exception cleft and must be flexible to allow entry and exit of made for the residues 33–45. Within this the polypeptide substrates and products, as well as particular region, Fwt PR has lower B-factor inhibitors. values as compared with Bwt, indicating lower Careful analysis of the flap hinges of both Bwt flexibility of the flap hinge region of the former and Fwt PRs shows that the loop containing enzyme. An equivalent behavior can be also residue 36 shifts toward the protein core in the clearly seen in the published structure of sub- Fwt structure and that the end of the side-chain type B multi-drug resistant variant containing in position 36 tends to occupy equivalent posi- the M36I mutation complexed with indinavir tions in both Bwt and Fwt in order to maintain and ritonavir.18 This indicates that the differences contacts with the interacting residues (Figure 3 observed in the flap hinge region of Fwt PR are (a)). This specific side-chain position and con- of a general nature and are not restricted to the formation of the residue 36 (Met or Ile) is binding of the particular inhibitor used in this maintained by van der Waals (VDW) interactions study. with the side-chains of neighboring residues It has been noted that the non-B subtypes HIV-PRs (Leu33, Leu38, and Lys20). In other words, do not usually develop the primary mutation L90M, mutation of a long methionine residue to a not even in the individuals under intensive treat- shorter isoleucine causes a collapse of the loop, ment with HIV-PR inhibitors, whereas the L89M 1034 Structural Characterization of B and F HIV PR

Figure 3. Pairwise comparisons of Bwt (red) with Fwt (in blue) and with Bmut (in green), emphasizing the effects of the M36I mutation. (a) The alignment of Bwt and Fwt shows a region of high discrepancy (circled) in the otherwise very similar structures, which cannot be seen in the alignment of Bwt and Bmut. (b) A close-up view of the flap hinges (on the right) reveals that this region contains the substitution M36I in subtype F HIV-PRs. Note that the position occupied by the end of the side-chain of residue 36 is the same in all three structures, forcing the backbone of a few neighboring amino acids in Fwt to shift toward the core of the protease, and thus increasing the number of van der Waals contacts between these residues and the rest of the protease. We propose that this displacement may be responsible by the hardening of the flap hinges, which, in turn, leads to higher susceptibility to development of resistance. Apparently the substitution S37A does not play a role in the observed shift, since it appears also in Bmut, where the structural difference is not significant. substitution is very frequent among the non-B The structural consequences of the L89M substitu- subtypes isolated.5 The double mutation L89M/ tion in the PRs that belong to the F subtype can be L90M is very rare in all subtypes. Here we propose related to those of the mutation L90M (Figure 5(b)). that the L89M substitution mimics the effect of the The longer methionine side-chain at position 89 L90M mutation, which could explain why they are forces the neighboring Leu90 to shift toward the mutually exclusive. Structurally, when the well- main chain atoms in the vicinity of the catalytic characterized mutation L90M occurs, the bulkier Asp25, mimicking the effects of the L90M mutation. side-chain of methionine makes additional van der This displacement increases the number of van der Waals contacts with the main chain atoms of the Waals interactions between the side-chain of Leu90 active site residues 24 through 26. The increased and the catalytic loop (five in Bwt to seven in Fwt/ number of interactions leads to a decrease of the Fmut), thus constraining the S1/S1′ binding pockets. volume of the substrate-binding cavity and com- The maximum shift observed for the active site promises the structural flexibility of the S1/S1′ residues is 0.3 Å for the carboxyl side-chain of Asp25 pockets, as compared with the wild-type protease. in the Fwt structure, which is in agreement with the This effect renders resistance to some commercial degree of structural rearrangements reported for the inhibitors, particularly nelfinavir and saquinavir. mutation L90M.19,20 This analysis, previously presented for the G48V/ Apart from M36I and L89M, all other amino acid L90M double mutant HIV-PR crystallized with sequence differences between Bwt and Fwt HIV-PRs saquinavir,19 can be extended to our Bmut structure are considered polymorphic. Five of those muta- complexed with TL-3, also containing the L90M tions, D60E, Q61N, I62V, L63S and I64L, are located mutation (Figure 5(a)). away from the active site, at the end of a β-strand b' Structural Characterization of B and F HIV PR 1035

Figure 4. Plot of the Fwt mean temperature factors per residue (Fwt B factor) relative to the Bwt B factor (in %) calculated as 100x(Fwt B factor)/(Bwt B factor) against the residue number (chain A). The mean B factors for the Fwt structure are consistently higher than those for the Bwt (gray shaded) except in a few black shaded areas. Within the region between residues 35 and 45, Fwt PR has lower B factor values as compared to Bwt,indicating lower flexibility of the flap hinge region of the former enzyme.

(nomenclature according to Wlodawer et al.21) which structure. At the S1′ side, however, the rotation of forms one side of the flap, while the substitution I15V the P1′ Phe side-chain decreases the number of occurs at the last residue of the β-strand b, in direct interactions between P1′ and S1′ (25/20 VDW contact with b'. The observed structural differences interactions in Bwt/Bmut and 39/19 VDW interac- are responsible for significantly lower inhibition of tions in Fwt/Fmut), leading to a loosely bound Fwt by TL-3 as compared to Bwt (Table 1). inhibitor and reduced binding affinity. Similar behavior was also observed for the symmetric Subtype B and F mutants inhibitor BMS-182193 in an HIV-PR containing the V82A mutation.23 This indicates that the asymmetry Both Bmut and Fmut contain the primary mutation between P1 and P1′ is caused by structural collapse V82A in the S1/S1′ pocket. The effect of this mutation of the S1′ pocket, and by rearrangements of the on the arrangement of the TL-3 inhibitor is similar to inhibitor's side-chain. It is also interesting to note that those previously reported22 for the symmetrical the rotation of the TL-3 P1 side-chain is one of the inhibitor A-77003. Structurally, only one of the two largest observed for this mutation. While for A-77003 hydroxyl groups of the central diol of TL-3 can occupy this difference is so small that it can be barely the middle position between the two catalytic noticed,22 for TL-3 it amounts to a rotation of 53° aspartate residues in the active site, forcing inhibitor between the Bwt/Bmut structures and 87° between binding to the protease active site in an asymmetric the Fwt/Fmut structures. manner. Such asymmetry forces the remaining The effects of the primary mutation L90M in Bmut hydroxyl to be positioned inside the S1′ pocket, as are equivalent to those described previously,19 shown in Figure 6. In the wild-type structures, the where the bulkier methionine residue shifts the side-chain of Phe in the P1 position occupies the active site loop containing the residues 24–26 central part of the pocket and interacts with the side- towards the S1/S1′ pockets, thus decreasing the chain of Val82 though van der Waals interactions. In enzyme flexibility, and ultimately leading to this arrangement, the P1′ Phe ring shifts away from reduced activity. Val82 due to the asymmetric binding of the inhibitor For two of the secondary mutations in the flap and occupies a crevice between the flap and the loop region, M46I (Fmut only) and I54V (both Bmut and containing the residue Pro81B (Figure 6). This Fmut), no significant structural differences could be displacement increases the number of interactions detected between the native and mutant enzymes. between the P1′ Phe side-chain and the S1′ pocket This is consistent with the hypothesis that these two when compared to P1/S1 (25 interactions on P1′/S1′ mutations affect inhibitor binding dynamically by against 14 interactions on P1/S1 for the Bwt struc- altering the flexibility of the flaps, but not the final ture). In the structures of Bmut and Fmut that contain closed conformation of the enzyme.18 the mutation V82A, the P1′ Phe side-chain rotates towards the void created by the mutation. This Consequences of polymorphic substitutions in rearrangement of the inhibitor's side-chain is caused non-B subtypes by a collapse of the flap residues and Pro81B, which seals the crevice in which it was inserted in the wild- It has been proposed that the presence of secondary type structures. All these rearrangements preserve the mutations in a virus infecting a drug-naïve individual number of van der Waals interactions observed could increase the risk of therapeutic failure.24 It was between P1 and S1 when compared to the Bwt shown that patients with PR mutation M36I were at 1036 Structural Characterization of B and F HIV PR

Figure 5. The structures of Bmut and Fwt PRs aligned on the subtype B wild-type PR (Bwt). Bwt is colored red, Bmut green, and Fwt blue. (a) Bwt compared to Bmut. (b) Bwt compared to Fwt. It appears that the polymorphic substitution L89M in the subtype F PRs, here represented by the Fwt PR, causes the same effect as the mutation L90M induced by anti- retroviral treatment in the subtype B proteases, as shown by Bmut PR. Both mutations displace the residue 90 toward the active site loop containing the residue Asp25, thus increasing the amount of van der Waals interactions between residue 90 and the active site loop. This shift slightly compresses the substrate/inhibitor binding cavity, leading to a reduction in volume and flexibility of the active site. significantly higher risk of having virologic failure at acute, since 99% of subtype A, 85% of subtype C, 69% week 24 when compared to patients carrying native of subtype F and 98% of subtype G naïve persons protease.24 According to the samples found in the carry this substitution as a consequence of natural Stanford HIV Database,25 98% of subtype A, 80% of polymorphism (subtypes B and D proteases show subtype C, 62% of subtype D, 95% of subtype F, and only 1% and 3% of the L89M substitution, respec- 100% of subtype G naïve individuals carry a tively). It appears, on the basis of these data, that the methionine rather than isoleucine at position 36 of subtype B HIV-PR is an exception among the other the protease, against only 13% of subtype B, which subtypes. leads to the conclusion that non-B HIV-PR subtypes The polymorphic substitutions L63P and A71V are more susceptible to develop anti-retroviral resis- present in Bmut, and L10I, K20R and T74A present in tance. If we consider that the substitution L89M is, at Fmut occur in persons receiving PIs therapy and, least in part, a structural equivalent of the primary together with other mutations, are associated with mutation L90M, then this susceptibility is even more resistance to each of the PIs.25 Finally, the proteins Structural Characterization of B and F HIV PR 1037

Figure 6. Effects of the mutation V82A on the inhibitor-binding mode. The inhibitor subsite P1/P1′ is shown in ball- and-stick representation along with the pocket S1/S1′ in stick representation, for the structures Bwt (red) and Bmut (green). The broken lines represent hydrogen bonds between the active site aspartate residues and the oxygen of the diol in the center of the inhibitor, whereas the asterisks mark residues that belong to chain B. Due to the asymmetric mode of binding of the inhibitor, the S1′ pocket, occupied by the P1′ phenyl side-chain, accommodates one hydroxyl of the central diol, which is indicated by a dotted circle. Whereas there is no significant modification in the P1 subsite due to V82A mutation (marked with a square), in the P1′ a rotation of the Phe ring of the inhibitor is observed. This rotation decreases the number of interactions between the inhibitor's side-chain and the S1′ pocket, while the interactions are maintained in the S1 pocket. discussed here have some other substitutions not and the accessory flap mutation I54V in Bmut and found in the Stanford HIV Database: K45R (Bmut), Fmut reduces the susceptibility to all PIs, notably S37A (Bmut and Fwt), R41K (Bmut, Fwt and Fmut), including indinavir, ritonavir, and lopinavir.25 If our Q61N/L63S/I64L (Fwt), I15V/E35D/R57K (Fwt and hypothesis that the substitution L89M is structurally Fmut) and G16E/S37N/P39S/G51R/Q61D/K70R/ equivalent to the primary mutation L90M is correct, I72V (Fmut). These findings call for care in the choice this would lead to intermediate-to-high-level resis- of inhibitors being used for treatment of patients tance of subtype F HIV-PR to all the commercial infected with non-B subtypes of HIV. protease inhibitors. The L90M mutation is most likely elicited due to administration of nelfinavir and Inhibitory assays saquinavir in the therapy regimen.26 In order to confirm these points, the affinities of commercial The inhibition constants (Ki) obtained for the inhibitors to the B and F subtype HIV-PRs described different protease subtypes against the inhibitor here are currently under investigation in our TL-3 are summarized in Table 1. The subtype F laboratory. Preliminary results obtained to date wild-type protease, as well as the B and F mutant clearly demonstrate higher resistance of Fwt, Fmut, proteases, present different degrees of resistance and Bmut PRs to the currently used commercial against TL-3. When compared with Bwt PR, the Ki inhibitors (our unpublished data). values are about threefold higher for Bmut PR, about sevenfold higher for Fwt PR, and about 54-fold higher for Fmut PR. The presence of the primary and Conclusions the accessory mutations in Fwt, Fmut and Bmut PRs, alone or together, as well as the described structural Mutations in the substrate cleft cause drug re- differences due to them, sheds light on the structural sistance by reducing the binding affinity between the basis of the decreased affinity of the TL-3 inhibitor to inhibitor and the mutant protease enzyme.27 Muta- PRs encoded by subtypes B and F of HIV-1. In tions elsewhere in the enzyme either compensate for addition, the presence of the primary mutation V82A the decreased kinetics of the enzymes with active site 1038 Structural Characterization of B and F HIV PR mutations or also cause resistance by altering enzyme sodium cacodylate (pH 6.2), 0.32 M ammonium sulfate, 6% catalysis, dimer stability, inhibitor binding kinetics, or MPD (condition 2)). Both Fwt and Fmut PRs were crystal- by re-shaping the active site through long-range lized using ammonium sulfate as precipitant. A decrease in structural perturbations.28 The presence of primary the crystallization temperature from 291 K to 277 K was and secondary drug-resistant mutations alone is not used to convert a shower of small rods to just a few small crystals per drop. The crystallization plate was then sufficient to explain the different degrees of resistance transferred back to 291 K to allow the crystals to reach to TL-3 inhibition of each HIV-PR discussed here. In their maximum size. It appears that the synergistic effect of order to understand these differences, the influence of a co-precipitant29 is a necessary condition to crystallize the the amino acid polymorphism on the binding of the Bmut PR, given the fact that no crystals have been obtained TL-3 inhibitor must be carefully considered. The using ammonium sulfate as the only precipitant. results presented here show that the polymorphic substitutions, which need not be in the active site, Data collection and crystal structure solution may amplify the effects of drug-resistant mutations. Therefore, the combined effects of naturally existing X-ray diffraction data for flash-cooled crystals of Bmut polymorphisms and drug-resistant mutations might (condition 2), Fwt, and Fmut were collected at the MX1 have important consequences on the viability of beamline of the Brazilian Synchrotron Light Laboratory current HIV-1 protease inhibitors with pharmaceu- (LNLS)30,31 using radiation with 1.43 Å wavelength, and a tical value. MarCCD detector. Crystals were frozen in the original crystallization solution containing 20% (v/v) glycerol. Initial diffraction of all protease crystals was significantly Materials and Methods improved by successive rounds of annealing, during which the frozen crystals were transferred to the cryo- solution at room temperature and then were flash-cooled Cloning and expression again. Measurements of data for the crystal of Bwt PR (Mar345 image plate detector mounted on a Rigaku RU- The cDNA of the whole virus genome from subtypes B 200 rotating anode) have been reported.11 mutant and both wild-type and mutant subtype F were The new datasets were integrated using MOSFLM 6.2.332 obtained from HIV-1 vertically infected seropositive and scaled with SCALA 3.1.20.33 The structure of Bmut Brazilian children.26 The patients infected with Bmut and protease was solved by molecular replacement using the Fmut viruses received treatment consisting of PR inhibi- structure of wild-type subtype B HIV-PR refined at 1.09 Å tors ritonavir/nelfinavir and reverse transcriptase inhibi- (PDB code: 1kzk) as the search model.34 The structure was tors amivudine/stavudine/zidovudine. initially determined with the program AmoRe35 in space The protease coding sequence was amplified by PCR, group P6122 (although identification of the space group as 9 simultaneously inserting the point mutation Q7K, and either P6122 or P61 was possible, as discussed above). The the insert was ligated in the pET11a vector that had been structures of the other proteases were solved in space 36 cleaved with the restriction enzymes NdeI and BamHI. groups P6122 and in P61 using the program PHASER 1.2, Escherichia coli BL21(DE3)-RIL (Stratagene) competent with the Bmut subunit serving as the search model. cells were transformed with each of the obtained clones Positional and temperature factor refinement were per- and the cells were cultivated at 37 °C in LB medium formed with CNS 1.137 and REFMAC 5.1.24.38 Water supplemented with 35 μg/ml chloramphenicol, 100 μg/ molecules were automatically inserted with ARP/ ml ampicilin, 1% (w/v) glucose and 0.05% antifoam 204 wARP.39 Real space refinement was carried out by visual (Sigma). The culture was induced with 1 mM IPTG when inspection of the model using the program O.40 TLS the A600nm reached 0.7. After 4 h of induction the cells were parameters were refined in the penultimate cycle of harvested by centrifugation. positional refinement, without application of non-crystallographic symmetry. Three TLS groups were defined for each subunit. The last cycle was performed re-establishing Inclusion bodies purification and refolding the non-crystallographic symmetry, with fixed TLS parameters. When appropriate, double side-chain conforma- The protocol for the production of the refolded Brazilian 15 tions were modeled. The electron density of the final models HIV-PRs has been reported. Briefly, the washed inclusion for all structures with the exception of Fmut clearly indicated bodies, dissolved in urea-containing buffer, were purified by chemical modification of the Cys67 amino acid residue. An two successive ion-exchange columns (Q Sepharose and SP additional electron density could be satisfactory modeled as Sepharose, respectively). The purified HIV-PR was refolded awholeβ-mercaptoethanol molecule covalently attached to by dialysis against a urea-free buffer, and concentrated for the cysteine residue (in one subunit of Bwt structure and in use in enzymatic assays and crystallization trials. both subunits of Bmut and Fwt models). β-Mercaptoethanol was used throughout the whole process from cell lysis to Crystallization refolding buffers. Three acetate anions, probably carried out from the cation exchange chromatography buffer, were identified on the surface of the Fwt structure, one in subunit The protease-inhibitor complexes were prepared by ∼ A and two in subunit B (none of them in the vicinity of the mixing the protease (at 0.4 mM) with twofold molar ex- inhibitor molecule). cess of TL-3 inhibitor (at 22 mM in DMF/DMSO 1:1, v/v) for 1 h on ice. X-ray diffraction quality crystals were obtained as described.15 Bmut PR crystals appeared after Enzymatic and inhibitory assays two days and continued to grow for about two weeks (one crystal per drop), in two conditions (0.1 M sodium The catalytic activities of the HIV-1 PRs, as well as the cacodylate (pH 6.2), 0.32 M ammonium sulfate, 6% (v/v) inhibition constants (Ki) for the inhibitor TL-3, were MPD, 5.1 % (w/v) PEG 3350 (condition 1), and 0.1 M monitored by following the hydrolysis of the fluorogenic Structural Characterization of B and F HIV PR 1039 substrate Arg-Glu(EDANS)-Ser-Gln-Asn-Tyr-Pro-Ile-Val- Supplementary Data Gln-Lys(DABCYL)-Arg (Molecular Probes). The proteases were added to a 300 μl cuvette containing substrate at 37 °C. Final concentrations in the standard assay were: 5–50 nM Supplementary data associated with this article protease, 10 μM substrate, 100 mM sodium acetate (pH 4.7), can be found, in the online version, at doi:10.1016/ 1 M sodium chloride, 1 mM EDTA, 1 mM DTT, 10% (v/v) j.jmb.2007.03.049 DMSO, plus increasing amounts of inhibitor. The fluorescence was monitored by using PC1 Photon Counting Spectrofluorometer (ISS) with excitation and emission References wavelengths of 340 nm and 490 nm, respectively. Different concentrations of EDANS in assay buffer 1. Vaishnav, Y. N. & Wong-Staal, F. (1991). The biochem- containing the substrate were used for calibration, con- istry of AIDS. Annu. Rev. Biochem. 60, 577–630. verting the fluorescence data sampled at regular time 2. Kuiken, C., Foley, B., Freed, E., Hahn, B., Marx, P., intervals into concentrations (nM). The reaction rates McCutchan, F., et al., (eds). (2002). HIV sequence −1 (nM×min ) were obtained from the slope of the product Compendium 2002. Published by Theoretical Biology concentration progress curves. In cuvettes, the initial and Biophysics Group, Los Alamos National Labora- – rate of wild-type or mutant PR was adjusted to 10 tory, LA-UR n 03-3564. −1 20 nM×min by addition of the proper amount of en- 3. Osmanov, S., Pattou, C., Walker, N., Schwardlander, – zyme. The reaction was followed for 8 10 min afterwards. B. & Esparza, J. (2002). Estimated global distribution For each inhibitor concentration, the steady-state rate in and regional spread of HIV-1 genetic subtypes in the the presence of inhibitor (Vi) was divided by the initial rate year 2000. J. Acquir. Immune Defic. Syndr. 29, 184–190. in the absence of inhibitor (V0). The resulting ratios Vi/V0 4. Kantor, R. & Katzenstein, D. (2003). Drug resistance in and the corresponding inhibitor concentrations were used non-subtype B HIV-1. J. Clin. Virol. 29, 152–159. to determine the inhibition constant (Ki) of the inhibitor 5. Caride, E., Hertogs, K., Larder, B., Dehertogh, P., and the total concentration of active enzyme (Et) by fitting Brindeiro, R., Machado, E. et al. (2001). Genotypic and the data to the general equation for competitive tight- phenotypic evidence of different drug resistance 41,42 binding inhibitors : mutation patterns between B and non-B subtype = isolates of human immunodeficiency virus type 1 Vi V0 ¼ð½Et½ItKi,app = found in Brazilian patients failing HAART. Virus 2 1 2 = 23 – þ fð½Et½ItKi,appÞ þ 4½EtKi,appg Þ 2½Et Genes, , 193 202. 6. Wlodawer, A. (2002). Rational approach to AIDS drug design through structural biology. Annu. Rev. Med. 53, where Vi is steady-state rate in the presence of the Δ −1 595–614. inhibitor, measured as [product]min ; V0 is initial rate in the absence of the inhibitor, measured as Δ 7. Ohtaka, H. & Freire, E. (2005). Adaptive inhibitors − 88 [product]min 1; [Et] is total concentration of the enzyme of the HIV-1 protease. Prog. Biophys. Mol. Biol. , – (free and bound); [It] is total concentration of the 193 208. 8. Shafer, R. W. (2002). Genotypic testing for human inhibitor (free and bound); Ki,app is apparent dissociation constant of the enzyme-inhibitor complex at a given immunodeficiency virus type 1 drug resistance. Clin. 15 – substrate concentration. Microbiol. Rev. , 247 277. Since under our experimental conditions the substrate 9. Rosé, J. R., Salto, R. & Craik, C. S. (1993). Regulation of concentration [S] is lower than the Michaelis-Menten autoproteolysis of the HIV-1 and HIV-2 proteases with 268 constant (K ), thus K and K are similar because engineered amino acid substitution. J. Biol. Chem. , m i,app i – competition of the inhibitor with the substrate is negli- 11939 11945. 10. Lee, T., Le, V.-D., Lim, D., Lin, Y.-C., Morris, G. M., gible. Ki,app corresponds to the IC50,theinhibitor concentration resulting in 50% inhibition of the enzymatic Wong, A. L. et al. (1999). Development of a new type of activity under defined experimental conditions. protease inhibitors, efficacious against FIV and HIV variants. J. Am. Chem. Soc. 121, 1145–1155. 11. Li, M., Morris, G. M., Lee, T., Laco, G. S., Wong, C. H., Protein Data Bank accession codes Olson, J. et al. (2000). Structural studies of FIVand HIV- 1 proteases complexed with an efficient inhibitor of The coordinates and structure factors have been FIV protease. Proteins: Struct. Funct. Genet. 38,29–40. submitted to the RCSB Protein Data Bank. The accession 12. Mildner, A. M., Rothrock, D. J., Leone, J. W., Bannow, codes are 2P3A, 2P3B, 2P3C and 2P3D for the Bmut, Bwt, C. A., Lull, J. M., Reardon, I. M. et al. (1994). The Fwt, and Fmut PRs, respectively. HIV-1 protease as enzyme and substrate: mutagen- esis of autolysis sites and generation of a stable mutant with retained kinetic properties. Biochemistry, 33, 9405–9413. 13. Murthy, K. H., Winborne, E. L., Minnich, M. D., Acknowledgements Culp, J. S. & Debouck, C. (1992). The crystal structures at 2.2- Å resolution of hydroxyethylene-based This work was supported in part by Fundação inhibitors bound to human immunodeficiency virus de Amparo à Pesquisa do Estado de São Paulo type 1 protease show that the inhibitors are present in two distinct orientations. J. Biol. Chem. 267, (FAPESP), Brazil, via grants 99/03387-4, 04/11890-8, 22770–22778. 04/12201-1 and 06/00182-8; by Conselho Nacional 14. Wlodawer, A. & Erickson, J. W. (1993). Structure- de Desenvolvimento Científico e Tecnológico based inhibitors of HIV-1 protease. Annu. Rev. (CNPq), Brazil; and by the Intramural Research Biochem. 62, 543–585. Program of the NIH, National Cancer Institute, 15. Sanches, M., Martins, N. H., Calazans, A., Brindeiro, Center for Cancer Research. R. M., Tanuri, A., Antunes, O. A. C. & Polikarpov, I. 1040 Structural Characterization of B and F HIV PR

(2004). Crystallization of a non-B and a B mutant HIV- 28. Muzammil, S., Ross, P. & Freire, E. (2003). A major role Protease. Acta Crystallog. sect. D, 60, 1625–1627. for a set of non-active site mutations in the develop- 16. Kervinen, J., Lubkowski, J., Zdanov, A., Bhatt, B., ment of HIV-1 protease drug resistance. Biochemistry, Dunn, B. M., Hui, K. Y. et al. (1998). Toward a universal 42, 631–638. inhibitor of retroviral proteases: comparative analysis 29. Majeed, S., Ofek, G., Belachew, A., Huang, C. C., of the interactions of LP-130 complexed with proteases Zhou, T. & Kwong, P. D. (2003). Enhancing protein from HIV-1, FIV, and EIAV. Protein Sci. 7, 2314–2323. crystallization through precipitant synergy. Structure, 17. Miller, M., Schneider, J., Sathyanarayana, B. K., Toth, 11, 1061–1070. M. V., Marshall, G. R., Clawson, L. et al. (1989). 30. Polikarpov, I., Teplyakov, A. & Oliva, G. (1997). The Structure of complex of synthetic HIV-1 protease with ultimate wavelength for protein crystallography? Acta a substrate-based inhibitor at 2.3 Å resolution. Science, Crystallog. sect. D, 53, 734–737. 246, 1149–1152. 31. Polikarpov, I., Perles, L. A., de Oliveira, R. T., 18. Clemente, J. C., Moose, R. E., Hemrajani, R., Whitford, Oliva, G., Castellano, E. E., Garratt, R. & Craievich, L. R., Govindasamy, L., Reutzel, R. et al. (2004). A. (1998). Set-up and experimental parameters of Comparing the accumulation of active- and nonac- the protein crystallography beamline at the Brazi- tive-site mutations in the HIV-1 protease. Biochemistry, lian National Synchrotron Laboratory. J. Synchrotron 43, 12141–12151. Rad. 5,72–76. 19. Hong, L., Zhang, X. C., Hartsuck, J. A. & Tang, J. 32. Leslie, A. G. W. (1992). Recent changes to the (2000). Crystal structure of an in vivo HIV-1 protease MOSFLM package for processing film and image mutant in complex with saquinavir: insights into plate data. Joint CCP4 and ESF-EACBM Newsletter on the mechanisms of drug resistance. Protein Sci. 9, Protein Crystallography, 26. 1898–1904. 33. Evans, P. R. (1997). In Proceedings of the CCP4 20. Kovalevsky, A. Y., Tie, Y., Liu, F., Boross, P. I., Wang, Study Weekend. Recent Advances in Phasing (Wilson, Y.-F., Leshchenko, S. et al. (2006). Effectiveness of K. S., Davies, G., Ashton, A. W. & Bailey, S., eds), nonpeptide clinical inhibitor TMC-114 on HIV-1 pp. 97–102, Daresbury Laboratory, Warrington. protease with highly drug resistant mutations D30N, 34. Reiling, K. K., Endres, N. F., Dauber, D. S., Craik, C. S. I50V, and L90M. J. Med. Chem. 49, 1379–1387. & Stroud, R. M. (2002). Anisotropic dynamics of the 21. Wlodawer, A., Miller, M., Jaskólski, M., Sathyanar- JE-2147-HIV Protease complex: drug resistance and ayana, B. K., Baldwin, E., Weber, I. T. et al. (1989). thermodynamic binding mode examined in a 1.09 Å Conserved folding in retroviral proteases: crystal structure. Biochemistry, 41, 4582–4594. structure of a synthetic HIV-1 protease. Science, 245, 35. Navaza, J. (1994). AMORE - An automated package 616–621. for molecular replacement. Acta Crystallog. sect. A, 50, 22. Baldwin, E. T., Bhat, T. N., Liu, B., Pattabiraman, N. & 157–163. Erickson, J. W. (1995). Structural basis of drug 36. Storoni, L. C., McCoy, A. J. & Read, R. J. (2004). resistance for the V82A mutant of HIV-1 proteinase. Likelihood-enhanced fast rotation functions. Acta Nature Struct. Biol. 2, 244–249. Crystallog. sect. D, 60, 432–438. 23. Kervinen, J., Thanki, N., Zdanov, A., Tino, J., Barrish, 37. Brünger, A. T., Adams, P. D., Clore, G. M., DeLano, J., Lin, P. F. et al. (1996). Structural analysis of the W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998). native and drug-resistant HIV-1 proteinases com- Crystallography and NMR system: a new software plexed with an aminodiol inhibitor. Protein Pept. suite for macromolecular structure determination. Letters, 3, 399–406. Acta Crystallog. sect. D, 54, 905–921. 24. Perno, C. A., Cozzi-Lepri, A., Balotta, C., Forbici, F., 38. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. (1997). Violin, M., Bertoli, A. et al. (2001). Secondary muta- Refinement of macromolecular structures by the tions in the protease region of human immunodefi- maximum-likelihood method. Acta Crystallog. sect. D, ciency virus and virologic failure in drug-naïve 53, 240–255. patients treated with protease inhibitor based therapy. 39. Perrakis, A., Morris, R. & Lamzin, V. S. (1999). J. Infect. Dis. 184, 983–991. Automated protein model building combined with 25. Rhee, S.-Y., Gonzales, M. J., Kantor, R., Betts, B. J., iterative structure refinement. Nature Struct. Biol. 6, Ravela, J. & Shafer, R. W. (2003). Human immunode- 458–463. ficiency virus reverse transcriptase and protease 40. Jones, T. A., Bergdoll, M. & Kjeldgaard, M. (1990). O: a sequence database. Nucl. Acids Res. 31, 298–303. macromolecular environment. In Crystallographic and 26. Brindeiro, P. A., Brindeiro, R. M., Mortensen, C., Modeling Methods in Molecular Design (Bugg, C. & Hertogs, K., Vroey, V. D., Rubini, N. P. M. et al. (2002). Ealick, S., eds), pp. 189–195, Springer-Verlag Press, Testing genotypic and phenotypic resistance in Berlin. Human Immunodeficiency Virus type 1 isolates of 41. Morrison, J. F. (1969). Kinetics of the reversible inhi- clade B and other clades from children failing bition of enzyme catalysed reactions of tight-binding antiretroviral therapy. J. Clin. Microbiol. 40, 4512–4519. inhibitors. Biochem. Biophys. Acta, 185, 269–285. 27. Dunn, B. M., Goodenow, M. M., Gustchina, A. & 42. Bieth, J. G. (1995). Theoretical and practical aspects of Wlodawer, A. (2002). Retroviral proteases. Genome proteinase inhibition kinetics. Methods Enzymol. 248, Biol. 3, 3006.1–3006.7. 59–84.

Edited by R. Huber

(Received 14 December 2006; received in revised form 14 March 2007; accepted 20 March 2007) Available online 24 March 2007